Research summary
Guardrails in Ontario: what 23 years of testing shows
Canadian Safety Group reviewed every published Canadian guardrail test we could find, from Ontario's own 2003 programme to Quebec's roof testing in 2021. The wood was rarely the problem. The fasteners were. And the rule does not mention them.
Five things the research shows
The findings in one screen. The rest of this page explains each one.
- Fasteners decide the strength. Guardrails built with common nails broke at a third of the legal load. The same design built with screws and two gussets held twice the legal load.
- You cannot tell by looking. A nailed guardrail and a screwed guardrail look the same. Some test guardrails looked well built and failed under a light push.
- A legal guardrail can still fail in a real fall. Quebec crash tests found that a bare, unbraced wood post is not strong enough. Ontario's rule allows exactly that post.
- Strong guardrails fail without warning. Weak nailed guardrails lean and lift slowly. Strong screwed guardrails hold much more, then snap all at once.
- Nobody has tested guardrails in Ontario since 2003. Ontario's own testers asked for more testing. It did not happen. The best Canadian evidence now sits in Quebec.
Why we looked
Falls kill more Ontario construction workers than any other hazard. Guardrails are the first line of defence: the law says use a guardrail first, and move to a harness only when a guardrail is not practical.
Most people assume that a guardrail built to the rule is a guardrail that holds. CSG wanted to know if the published testing agreed. We went back to every Canadian test report we could locate, plus the current European standard, and compared what they found against what Ontario's regulation requires.
This page is the plain-language version of that review. The full position paper and findings summary, with every citation, are listed at the end.
What the rule says
O. Reg. 213/91 (Construction Projects), section 26.3.
A wood guardrail in Ontario must be built from spruce, pine or fir, construction grade or better, with no visible defects. Posts must be at least 38 x 89 mm (a 2x4), "securely fastened" to the surface, and no more than 2.4 m (8 ft) apart. The top rail sits 0.9 m to 1.1 m (35 in to 43 in) above the surface, with a mid rail halfway and a toe board at least 89 mm (3.5 in) high.
It must also hold four separate loads without failing. On site, think of them in pounds:
| Where | Load in the rule | In plain terms |
|---|---|---|
| Top rail, sideways | 675 N | About 150 lb |
| Top rail, straight down | 450 N | About 100 lb |
| Mid rail | 450 N | About 100 lb |
| Toe board, sideways | 225 N | About 50 lb |
Two things the rule does not say. It does not require the post to be braced or reinforced in any way. And it does not name a fastener type, size or count at any joint. "Securely fastened" is the whole instruction.
The testing below is about those two gaps.
Ontario, 2003: more than 90 guardrails pulled to failure
Construction Safety Association of Ontario, Test Report: 2x4 Wood Guardrail Assembly, October 2003.
The Construction Safety Association of Ontario (now part of IHSA) built wood guardrails on formply and OSB decks and pulled the top rail sideways until each one broke. They changed one thing at a time: the type of nail or screw, whether the post had a plywood gusset, how many gussets, and how the post was turned.
The chart shows the average breaking load of the end post in each build. The end post is always the weakest post in a run, and the one a worker is most likely to land against.
Average end-post breaking load, CSAO 2003. Middle posts in the same runs held 15 to over 50 per cent more, because the rails pass load sideways to the next post. Individual results in one setup ranged from about 120 lb (535 N) to 330 lb (1,470 N): wood is not consistent.
What the numbers mean
- Same wood, same size, same shape. Only the fasteners and gussets changed. The result moved from 45 lb (200 N) to 298 lb (1,325 N).
- Screws beat spiral nails. Spiral nails beat common nails. Every time.
- Two gussets beat one. Adding a second gusset lifted a fully screwed build from 164 lb (730 N) to 298 lb (1,325 N).
- One weak joint sinks the post. Screws everywhere except a spiral-nailed gusset still failed at 112 lb (500 N).
The test crew also reported two things that matter as much as the numbers. Several guardrails looked sound and well built, then broke under a light push. And the failures had two very different characters: nailed bases let go gradually, with the post leaning and the base lifting, while screwed bases held far more and then failed suddenly, often by snapping the post.
CSAO closed its report by recommending more testing, and asked other provinces to share data for comparison. As far as CSG can find, no Ontario body has tested a guardrail since.
Quebec, 2009: someone finally tested a real fall
Lan and Daigle, Safety Science, 2009. Funded by the IRSST, Quebec's occupational health and safety research institute.
Every number on the chart above came from a slow, steady pull. A worker tripping into a rail is neither slow nor steady. Quebec researchers built a test for that: a 100 kg (220 lb) wooden torso on a rolling frame, hitting the top rail at 2 m per second (6.5 ft/s). That is roughly walking speed, with a full-sized body behind it.
The finding
The crash test showed that a plain 2x4 post, on its own, is not strong enough to stop a falling worker. The post needs help: plywood gussets at the base, or a second 2x4 beside it.
Ontario's rule allows the plain post, with no help. So a guardrail can pass the rule and still be the post that failed this test.
"Does it meet the rule?" and "Will it stop a falling worker?" are two different questions. This test is where they came apart.
Quebec, 2019 to 2021: a much bigger programme
Galy and Lan, IRSST Report R-1048 (2019) and Practice Periodical on Structural Design and Construction (2020); Galy (2021).
The IRSST then ran more than 350 lab tests on job-built 2x4 guardrails fastened the way they actually get fastened on a house: to open-web joists and to prefabricated walls. They added 36 tests on real flat roofs. The results:
- Screws and lag screws beat nails. The same answer Ontario got in 2003, on a far bigger sample.
- The worst push is straight onto a post, not midway between posts. That is the case to build for.
- Shorter guardrails held more. Guardrails built at 1 m (3.3 ft) outperformed those at 1.2 m (4 ft). Ontario allows up to 1.1 m (43 in) without distinction.
- A steel angle bracket with plain nails was a real improvement over nails alone. Screws were still better.
- On existing flat roofs, not one steel guardrail base fixed to a single face of a parapet reached the required load. Zero out of 36. Bases that gripped both faces of the parapet did. Fastener count mattered more than fastener type.
The parapet finding matters for every re-roofing job. From above, a sound parapet and a rotten one look the same, and a guardrail fastened to one face of it looks finished. The testing says it is not.
Europe, 2025: a different way of checking
EN 13374:2025, Temporary edge protection systems.
CSG also looked at how Europe verifies temporary guardrails, not because the loads are higher (on point load, Ontario's 675 N (150 lb) is above the European 0.30 kN (67 lb) serviceability load for flat surfaces) but because the method is different. The European standard:
- puts a deflection limit on the load, so a rail that bends too far fails even if it does not break;
- tests the whole assembled system, not just its parts;
- for classes used on sloped surfaces, requires a drop-weight impact test, measured as energy absorbed rather than force resisted;
- sets a minimum principal guardrail height of 1.0 m (39 in) at any point.
One honest limit: the European class for flat surfaces (Class A) is static only, like Ontario. The European standard does not by itself prove that flat-surface guardrails need a dynamic test. The Canadian crash testing does that. What Europe shows is the direction verification has moved, and two things Ontario's rule has never had: a deflection limit and a whole-system check.
Note: the EN 13374:2025 values above were checked against the full text of the standard in August 2026 and verified by a competent person.
The 23-year gap
The gap matters because the 2003 tests were static only, used wood decks only, and were run under Ontario's rule. Quebec's rules are stricter, so its findings do not map onto Ontario exactly. Ontario is relying on a 23-year-old test programme that asked for a follow-up it never got.
Why this matters now
Two points follow from the Coroner's 2025 Construction Death Review. First, the workers most at risk are on small residential crews, and those are the crews most likely to build a guardrail on site from lumber rather than buy a manufactured system. Second, training was not the missing piece. The gap was between what was trained and what was built.
Training, assessor checklists and inspection all teach the same thing: check that the right parts are there, at the right heights and spacing. That is a fair reading of the rule. But the one variable every study calls decisive, the fasteners, is the one thing that check cannot see. A worker, an assessor or an inspector can do everything they were taught and still approve a guardrail that fails in the real event.
What we are NOT saying
We are not saying guardrails fail. A well-built guardrail holds, and the tests prove that too.
We are not telling anyone to ignore the rule. The rule is the law. Build to it.
Our whole point is two lines:
- The rule allows one weak build - a bare post with no bracing. Testing found that build can fail in a real fall.
- The rule never names the fasteners. In every test, the fasteners decided whether the guardrail held.
So build to the rule, then build past it: screws, two gussets, posts closer together. That guardrail passes the rule and passes the testing.
What you can do today
None of this needs a rule change. All of it comes from the test results or current IHSA guidance.
- Screws at all three base joints. Gusset to post, foot to base plate, base plate to deck. No exceptions.
- Two gussets on every post. The cheapest strength gain in the whole test set. Never build a bare post just because the rule allows it.
- Build at 0.9 m (35 in), not 1.1 m (43 in). Shorter guardrails held more.
- Bring posts closer than 2.4 m. IHSA suggests 1 m to 2 m.
- On roofs, grip both faces of the parapet and use every fastener hole the manufacturer specifies.
- Inspect at the fastener, not from three metres away. Get close. Nails at the base mean the guardrail is not there: barricade, tie off, rebuild with screws.
- Never pull one rail and leave the posts. Rebuild the same day with the same type and number of fasteners. Rebuilds are where the nails creep back in.
Our Guardrail Field Guide (CSG-GDE-GRD-001) and Roof and Parapet Guide (CSG-GDE-GRD-002, coming to this page) cover the build details with drawings and a daily check. Both work on a phone.
How this review was done, and its limits
- CSG reviewed the published test reports and peer-reviewed papers listed below, the current text of O. Reg. 213/91, the IHSA manual, EN 13374:2025, and WSIB and Chief Coroner data.
- All Ontario data comes from one 2003 programme on wood decks. Concrete, steel and precast were not tested.
- Quebec's work was done under Quebec's rules, which are stricter than Ontario's.
- Wood varies. Moisture, knots and rot all change results and none can be seen from a distance.
This is general information for Ontario construction workplaces. It is not legal advice, and it is not an engineering opinion on any specific guardrail. Where a question is close, read the regulation and get advice from a competent person.
Sources
Links go to free official copies where one exists. Journal papers open at the publisher and may need library access.
- Construction Safety Association of Ontario (2003). Test Report: 2x4 Wood Guardrail Assembly. Technical Services and Quality Assurance, 30 October 2003. Not published online - contact us and we will point you to it.
- Lan, A. and Daigle, R. (2009). Development and validation of a method for evaluating temporary wooden guardrails built and installed on construction sites. Safety Science 47(2).
- Galy, B. and Lan, A. (2019). Resistance des garde-corps en bois fixes sur des structures neuves et comportement sous charges de garde-corps metalliques fixes sur des structures existantes. IRSST Report R-1048. Open access, free to download (French).
- Galy, B. and Lan, A. (2020). Strength of Temporary Wooden Guardrails Installed on Open Joists and Prefabricated Walls. Practice Periodical on Structural Design and Construction 25(2).
- Galy, B. (2021). In Situ Strength Testing of Temporary Steel Guardrails Installed on Existing Flat-Roof Parapets. Practice Periodical on Structural Design and Construction 26(1).
- EN 13374:2025. Temporary edge protection systems: product specification, test methods. CEN, May 2025. Sold through national standards bodies; not free online.
- Ontario Regulation 213/91, Construction Projects, sections 26, 26.1 and 26.3, as currently in force. Free on ontario.ca.
- Infrastructure Health and Safety Association. Construction Health and Safety Manual, Chapter 18: Guardrails.
- Office of the Chief Coroner for Ontario (2025). Construction Death Review: Falls from Heights. Free PDF.
- Workplace Safety and Insurance Board, claims data 2019 to 2023.
- Canadian Safety Group Inc. Position paper CSG-POS-2026-GRD-001 and findings summary CSG-BRF-2026-GRD-001, available on request.